Predicting the probability distribution of ice load amplitudes on ship hull in different ice and operational conditions

Mikko Kotilainen

Predicting the probability distribution of ice load amplitudes on ship hull in different ice and operational conditions

Julkaisun otsikon käännös: Laivan runkoon kohdistuvan jääkuormajakauman ennustaminen eri jää- ja operointiolosuhteissa

Mikko Kotilainen

Tutkimustuotos: Doctoral Thesis › Collection of Articles

Abstrakti

Suurten tavarankuljetuskapasiteettiensa ansiosta laivat ovat ympäristöystävällinen ja taloudellinen tavarankuljetusmuoto. Talvella laivojen täytyy kulkea jääpeitteisissä vesissä, mikä tekee talvimerenkulusta tärkeän insinööritieteen haaran. Jotta ympäristöä ja ihmishenkiä voidaan turvata jäissä navigoinnin aikana, meidän täytyy ymmärtää kuinka suuria kuormia jäissä navigointi aiheuttaa laivan runkoon. Kuitenkin monet epävarmuudet jään mikrorakenteessa sekä jäänmurtoprosessissa vaikeuttavat kuormien determinististä ennustamista, ja täten ennustaminen tehdään tilastollisilla menetelmillä. Tässä työssä analysoidaan täysmittakaavan jääkuormamittausaineistoa ja kuorman tapahtuessa vallinneita olosuhteita, saadaksemme selville miten jääkuormajakauma muuttuu eri olosuhteissa. Selvittääksemme miten kuormajakauman parametrit ja niihin liittyvät epävarmuudet muuttuvat olosuhteiden mukana, työssä käytettiin hierarkkisia Bayesiläisiä malleja, joille annettiin priorit Gaussisista prosesseista. Jäänpaksuus mallinnettiin Gaussisella prosessilla, jolle annettiin Studentin t-jakautunut virhetermi, kun taas laivan nopeus mallinnettiin lineaarisella interpoloinnilla. Erilaisten jääkuormamallien vertailu tehtiin vertaamalla mallien posterioreiden ennusteita mittauksiin, joissa Weibull-malli suoriutui parhaiten. Mallien mukaan jääkuormatasot muuttuivat olosuhteiden mukaan, nousten paksummassa jäässä, kun taas nopeuden vaikutus kuormiin oli vaikeaselkoisempi eri malleissa. Sen lisäksi, että kuormia analysoitiin yhdellä keulakaarella, kolmen peräkkäisen kaaren analysointi keulaolkapäässä mahdollisti paikallisten huippukuormien tunnistamisen. Tunnistamisen jälkeen analysoimme stereokameran kuvia huippukuorman syntymisen ajalta, kategorisoiden huippukuormat jään vasteen mukaan, selvittääksemme miten jään vaste vaikuttaa huippukuormien jakaumaan. Tämän jälkeen tutkittiin tarkemmin tapauksia, joissa huippukuorman aiheutti keulaolkapäässä kääntyvä jäälohkare, selvittäen miten jäälohkareen pinta-ala riippui jäänpaksuudesta ja laivan nopeudesta, sekä miten huippukuorma riippui lohkareen koosta ja laivan nopeudesta. Jäälohkareen pinta-ala oli suurempi hitaammilla nopeuksilla ja paksummalla jäällä, kun taas huippukuormatasot olivat suuremmat hitaammilla nopeuksilla ja suuremmilla lohkareilla.

Julkaisun otsikon käännös	Laivan runkoon kohdistuvan jääkuormajakauman ennustaminen eri jää- ja operointiolosuhteissa
Alkuperäiskieli	Englanti
Pätevyys	Tohtorintutkinto
Myöntävä instituutio	Aalto-yliopisto
Valvoja/neuvonantaja	Suominen, Mikko, Vastuuprofessori Vanhatalo, Jarno, Ohjaaja Kujala, Pentti, Ohjaaja
Kustantaja	Aalto University
Painoksen ISBN	978-952-64-1797-4
Sähköinen ISBN	978-952-64-1798-1
Tila	Julkaistu - 2024
OKM-julkaisutyyppi	G5 Artikkeliväitöskirja

Tutkimusalat

täysmittakaava
jääkuorma
jääolosuhteet
Bayesiläinen hierarkkinen malli
Gaussinen prosessi

Pääsy asiakirjaan

http://urn.fi/URN:ISBN:978-952-64-1798-1

Siteeraa tätä

@phdthesis{5a0366bab6284511a84f25df47a5aec4,

title = "Predicting the probability distribution of ice load amplitudes on ship hull in different ice and operational conditions",

abstract = "Due to their large cargo carrying capacities, ships are an environmentally and economically efficient mode for transporting cargo. During winters, the ships must navigate in ice-prone waters and hence shipping in ice-covered waters is an important engineering topic. To protect human lives and the environment during the shipping operations, we must understand the forces the ship hull sustains during the shipping operations. However, because of many uncertainties in the ice microstructure, and the ice breaking process, the forces required for breaking the ice cannot be resolved deterministically, but rather statistically. This work analyzed full-scale ice load data and the measured covariate conditions at the instant the load occurred to examine how the load distribution changes in different conditions. The work used Bayesian hierarchical models with Gaussian process priors to infer how the load probability distribution parameters and their uncertainties change as a function of condition covariates. The ice thickness was modeled using a Gaussian process model with a Student-t error model, whereas the ship speed was interpolated linearly. Different ice load models were compared via their posterior predictions, and the Weibull model was found to have the best posterior predictions. According to the trained models, the ice load levels were changing with conditions, increasing in thicker ice, whereas the speed effect was more convoluted across models. In addition to analyzing the loads from one bow frame, the research was able to identify local temporal maxima of loads by analyzing force measurements from three adjacent frames in the bow shoulder area. After identifying the peak loads from the data, the stereo camera photos were analyzed at the peak load occurrence times to study how the ice response affected the peak load distribution. The work investigated further the cases where the peak load was caused by a rotating ice cusp, studying how the cusp area was affected by the ice thickness and ship speed, and how the peak load was affected by the cusp's size and ship speed. The cusp area increased in slower speeds and thicker ice, whereas the peak loads increased with slower speeds and larger cusps.",

keywords = "full-scale, ice load, ice conditions, Bayesian hierarchical model, Gaussian process, t{\"a}ysmittakaava, j{\"a}{\"a}kuorma, j{\"a}{\"a}olosuhteet, Bayesil{\"a}inen hierarkkinen malli, Gaussinen prosessi, full-scale, ice load, ice conditions, Bayesian hierarchical model, Gaussian process",

author = "Mikko Kotilainen",

year = "2024",

language = "English",

isbn = "978-952-64-1797-4",

series = "Aalto University publication series DOCTORAL THESES",

publisher = "Aalto University",

number = "93",

address = "Finland",

school = "Aalto University",

}

TY - BOOK

T1 - Predicting the probability distribution of ice load amplitudes on ship hull in different ice and operational conditions

AU - Kotilainen, Mikko

PY - 2024

Y1 - 2024

N2 - Due to their large cargo carrying capacities, ships are an environmentally and economically efficient mode for transporting cargo. During winters, the ships must navigate in ice-prone waters and hence shipping in ice-covered waters is an important engineering topic. To protect human lives and the environment during the shipping operations, we must understand the forces the ship hull sustains during the shipping operations. However, because of many uncertainties in the ice microstructure, and the ice breaking process, the forces required for breaking the ice cannot be resolved deterministically, but rather statistically. This work analyzed full-scale ice load data and the measured covariate conditions at the instant the load occurred to examine how the load distribution changes in different conditions. The work used Bayesian hierarchical models with Gaussian process priors to infer how the load probability distribution parameters and their uncertainties change as a function of condition covariates. The ice thickness was modeled using a Gaussian process model with a Student-t error model, whereas the ship speed was interpolated linearly. Different ice load models were compared via their posterior predictions, and the Weibull model was found to have the best posterior predictions. According to the trained models, the ice load levels were changing with conditions, increasing in thicker ice, whereas the speed effect was more convoluted across models. In addition to analyzing the loads from one bow frame, the research was able to identify local temporal maxima of loads by analyzing force measurements from three adjacent frames in the bow shoulder area. After identifying the peak loads from the data, the stereo camera photos were analyzed at the peak load occurrence times to study how the ice response affected the peak load distribution. The work investigated further the cases where the peak load was caused by a rotating ice cusp, studying how the cusp area was affected by the ice thickness and ship speed, and how the peak load was affected by the cusp's size and ship speed. The cusp area increased in slower speeds and thicker ice, whereas the peak loads increased with slower speeds and larger cusps.

AB - Due to their large cargo carrying capacities, ships are an environmentally and economically efficient mode for transporting cargo. During winters, the ships must navigate in ice-prone waters and hence shipping in ice-covered waters is an important engineering topic. To protect human lives and the environment during the shipping operations, we must understand the forces the ship hull sustains during the shipping operations. However, because of many uncertainties in the ice microstructure, and the ice breaking process, the forces required for breaking the ice cannot be resolved deterministically, but rather statistically. This work analyzed full-scale ice load data and the measured covariate conditions at the instant the load occurred to examine how the load distribution changes in different conditions. The work used Bayesian hierarchical models with Gaussian process priors to infer how the load probability distribution parameters and their uncertainties change as a function of condition covariates. The ice thickness was modeled using a Gaussian process model with a Student-t error model, whereas the ship speed was interpolated linearly. Different ice load models were compared via their posterior predictions, and the Weibull model was found to have the best posterior predictions. According to the trained models, the ice load levels were changing with conditions, increasing in thicker ice, whereas the speed effect was more convoluted across models. In addition to analyzing the loads from one bow frame, the research was able to identify local temporal maxima of loads by analyzing force measurements from three adjacent frames in the bow shoulder area. After identifying the peak loads from the data, the stereo camera photos were analyzed at the peak load occurrence times to study how the ice response affected the peak load distribution. The work investigated further the cases where the peak load was caused by a rotating ice cusp, studying how the cusp area was affected by the ice thickness and ship speed, and how the peak load was affected by the cusp's size and ship speed. The cusp area increased in slower speeds and thicker ice, whereas the peak loads increased with slower speeds and larger cusps.

KW - full-scale

KW - ice load

KW - ice conditions

KW - Bayesian hierarchical model

KW - Gaussian process

KW - täysmittakaava

KW - jääkuorma

KW - jääolosuhteet

KW - Bayesiläinen hierarkkinen malli

KW - Gaussinen prosessi

KW - full-scale

KW - ice load

KW - ice conditions

KW - Bayesian hierarchical model

KW - Gaussian process

M3 - Doctoral Thesis

SN - 978-952-64-1797-4

T3 - Aalto University publication series DOCTORAL THESES

PB - Aalto University

ER -

Predicting the probability distribution of ice load amplitudes on ship hull in different ice and operational conditions

Abstrakti

Tutkimusalat

Pääsy asiakirjaan

Sormenjälki

Siteeraa tätä